English Français العربية

Which AI model should you use?

Prices, context windows, release dates and lifecycle for every major AI model, from chat to image, video and voice. Checked against each provider's own pages.

models
276
providers
13
With prices
163
Categories
12

Today's models

Current models by category, with their published prices and recent price moves.

All Moderation
# Model Price Released Price change
1
Free September 2024
2
No data March 2026
3
Shieldstral 1.0 Preview Mistral AI
No data No data

Just released

Price changes

  • Gemini 4 Argon Announced Input $2.00 → $4.00 · Output $10.00 → $20.00
    ▲ 100%
    from Dec 1
  • Gemini 3.5 Flash-Lite Announced Input $0.30 → $0.60 · Output $2.50 → $5.00
    ▲ 100%
    from Jan 1, 2027
  • Gemini 3.8 Flash Announced Output $3.75 → $7.50 · Input $0.75 → $1.50
    ▲ 100%
    from Jan 1, 2027
  • Gemini 3.8 Flash TTS Announced Input (text) $0.50 → $1.00 · Audio output (audio) $9.00 → $18.00
    ▲ 100%
    from Jan 1, 2027
  • Gemini 3.8 Flash-Lite TTS Announced Audio output (audio) $6.00 → $12.00 · Input (text) $0.50 → $1.00
    ▲ 100%
    from Jan 1, 2027
  • Gemini Omni Flash Announced Input $1.50 → $3.00 · Output (video) $17.50 → $35.00 · Output $9.00 → $18.00
    ▲ 100%
    from Jan 1, 2027

Amber: announced by the provider, not yet in force. Green or red: already applied.

Full timeline

Retiring soon

The most capable models right now

Artificial Analysis Intelligence Index, independent and run the same way for every model. Current language models, best reasoning setting tested.

Methodology
Intelligence Index, top 12
  1. Claude Opus 5.5Anthropic · $8.00 / 1M tokens57.6
  2. Claude Sonnet 5.5Anthropic · $4.00 / 1M tokens56.0
  3. Muse Wonder FullMeta55.8
  4. Claude Fable 5.1Anthropic · $20.00 / 1M tokens53.4
  5. GPT-6 AstraOpenAI · $20.00 / 1M tokens52.7
  6. Gemini 4 ArgonGoogle · $4.00 / 1M tokens52.6
  7. Muse WonderMeta52.4
  8. GPT-6.1 SolOpenAI · $4.00 / 1M tokens51.8
  9. Muse Spark 1.3Meta48.1
  10. GPT-6 SolOpenAI · $4.00 / 1M tokens47.6
  11. GPT-5.6 SolOpenAI · $8.00 / 1M tokens47.0
  12. Grok 4.7xAI · $3.00 / 1M tokens46.4
Most capable per dollarIndex points per dollar of blended price (per million tokens), among models scoring at least 30.
  1. Claude Haiku 5.5Anthropic · 43.4 / $0.20217
  2. GPT-6 LunaOpenAI · 38.1 / $0.20191
  3. DeepSeek V4.1 FlashDeepSeek · 45.1 / $0.5386
  4. GPT-5.6 LunaOpenAI · 37.3 / $0.4583
  5. Mistral Large 4Mistral AI · 38.4 / $0.7551
  6. Gemini 3.8 FlashGoogle · 40.9 / $1.5027
  7. DeepSeek V4 ProDeepSeek · 36.0 / $1.9818
  8. Grok 4.7xAI · 46.4 / $3.0015
  9. Claude Sonnet 5.5Anthropic · 56.0 / $4.0014
  10. Gemini 4 ArgonGoogle · 52.6 / $4.0013
  11. GPT-6.1 SolOpenAI · 51.8 / $4.0013
  12. GPT-6 SolOpenAI · 47.6 / $4.0012

Not sure which model fits?

Answer six questions and get three recommendations, with the reasoning behind each. No sign-up.

Find my model

The market at a glance

Current language models only, so like is compared with like.

The 15 lowest pricesBlended at 3 input tokens for every output token, cheapest first.
  1. DeepSeek V4 Flash 0731DeepSeek$0.15
  2. Claude Haiku 5.5Anthropic$0.20
  3. GPT-6 LunaOpenAI$0.20
  4. GPT-6 Luna ProOpenAI$0.20
  5. DeepSeek V4 Flash Vision ExpDeepSeek$0.32
  6. GPT-5.6 LunaOpenAI$0.45
  7. DeepSeek V4.1 FlashDeepSeek$0.53
  8. Mistral Large 4Mistral AI$0.75
  9. Qwen3.8 27BAlibaba Cloud (Qwen)$0.78
  10. Gemini 3.5 Flash-LiteGoogle$0.85
  11. Grok Build 0.1xAI$1.25
  12. Gemini 3.8 FlashGoogle$1.50
  13. Gemini 3.7 FlashGoogle$1.50
  14. Grok 4.3xAI$1.56
  15. Grok 4.20 ReasoningxAI$1.56

Models without a published price are left out rather than drawn as free.

Price range by providerEach dot is one model. Cheapest to dearest, blended rate. Logarithmic scale: each step is ten times the last.
  1. DeepSeek$0.15–$1.98
  2. Anthropic$0.20–$20.00
  3. OpenAI$0.20–$20.00
  4. Mistral AI$0.75
  5. Alibaba Cloud (Qwen)$0.78–$6.00
  6. Google$0.85–$4.50
  7. xAI$1.25–$3.00
  8. Moonshot AI$4.00
Show data
ModelValue
DeepSeek · DeepSeek V4 Pro$1.98
DeepSeek · DeepSeek V4.1 Flash$0.53
DeepSeek · DeepSeek V4 Flash 0731$0.15
DeepSeek · DeepSeek V4 Flash Vision Exp$0.32
Anthropic · Claude Fable 5.1$20.00
Anthropic · Claude Opus 5.5$8.00
Anthropic · Claude Sonnet 5.5$4.00
Anthropic · Claude Haiku 5.5$0.20
Anthropic · Claude Mythos 5.1$20.00
OpenAI · GPT-6 Astra$20.00
OpenAI · GPT-6.1 Sol$4.00
OpenAI · GPT-6 Luna$0.20
OpenAI · GPT-6 Sol$4.00
OpenAI · GPT-5.6 Sol$8.00
OpenAI · GPT-5.6 Terra$4.50
OpenAI · GPT-5.6 Luna$0.45
OpenAI · Chat Latest$11.25
OpenAI · GPT-6 Astra Pro$20.00
OpenAI · GPT-6 Luna Pro$0.20
OpenAI · GPT-6 Sol Pro$4.00
OpenAI · GPT-6.1 Sol Pro$4.00
OpenAI · GPT-6.1 Terra$2.00
OpenAI · GPT-6.1 Terra Pro$2.00
Mistral AI · Mistral Large 4$0.75
Alibaba Cloud (Qwen) · Qwen3.8 2.4T A95B$3.00
Alibaba Cloud (Qwen) · Qwen3.8 27B$0.78
Alibaba Cloud (Qwen) · Qwen3.8 Max Prime$6.00
Google · Gemini 4 Argon$4.00
Google · Gemini 3.8 Flash$1.50
Google · Gemini Omni Flash$3.38
Google · Gemini 3.5 Flash-Lite$0.85
Google · Gemini 3.7 Flash$1.50
Google · Gemini 3.1 Pro$4.50
xAI · Grok 4.7$3.00
xAI · Grok 4.3$1.56
xAI · Grok 4.20 Reasoning$1.56
xAI · Grok 4.20 Non-Reasoning$1.56
xAI · Grok 4.20 Multi-Agent$1.56
xAI · Grok Build 0.1$1.25
Moonshot AI · Kimi K3.1 Pro$4.00

Explore

Cost calculator

Enter your monthly usage and see what every model would cost.

Compare models

Put up to four models side by side, or read head-to-head pages.

Timeline

Releases, price changes, deprecations and retirements, as they happen.

AI tools

The apps built on these models: assistants, coding, image, video, voice.

AI glossary

146 terms explained in English, French and Arabic.

API status

Is a provider's API up right now? Live from their status pages.

Frequently asked questions

What is AI.tn?

A comparison of 276 AI models from 13 providers: language, image, video, voice, embedding and more. For each model it shows the published price, context window, release date, capabilities and whether it is current, legacy, deprecated or retired.

Where do the prices come from?

From each provider's own pricing and documentation pages, never from third-party aggregators. Every price records its source and the date it was checked. When a provider does not publish a price, the site says "No data" rather than estimating.

How are prices compared across models?

Language models are priced per million tokens, separately for input and output. To rank them with one number, the site uses a blended rate of three input tokens per output token. Other categories use their own unit: per image, per second of video, per minute of audio. Prices are only compared within a category.

What do legacy, deprecated and retired mean?

Legacy models still work but have newer replacements. Deprecated models have an announced shutdown date. Retired models no longer respond and are kept for reference with their last published price. The recommendation wizard only suggests current models.

Does the recommendation wizard use AI?

No. It scores models from published data with fixed, documented rules, so the same answers always give the same result and every recommendation shows why it ranked where it did. The method is explained on the methodology page.

Can I reuse the data?

There is a free read-only JSON API at /api/models, and an RSS feed of price changes. Always confirm a price on the provider's own page before committing to a model.